/srv/irclogs.ubuntu.com/2011/01/20/#ubuntu-installer.txt

svdaseinI'm having a bit of trouble getting 10.04's installer working right with the  serial console on my soekris board here.  Anyone here have any experience w/ that and/or can recommend another group?04:36
arahello!09:20
araI am trying to install ubuntu server (natty daily) in some servers with preseed, but in most of the systems it is stuck because it asks for a partition table to use09:21
aracould it be that the prior installation (also natty) corrupted the partition table somehow?09:22
arathis is the preseed file that I am using: http://paste.ubuntu.com/556098/09:24
JemtNot sure this is the place to ask, but this channel seems to be the best place to find technical assistance.11:10
JemtI'm remastering Ubuntu. One of my very last steps before creating my new ISO, is upgrading all packages (apt-get upgrade). Unfortunately this results in a lot of errors/warnings from dpkg: "dpkg: error processing gconf2-common (--configure): dependency problems - leaving unconfigured". Another common error is this: "dpkg: dependency problems prevent configuration of compiz-gnome: compiz-gnome depdends on libgconf2-4 (>= 2.31.1); however:11:13
Jemtpackage libgconf2-4 is not configured yet". Are these just warnings, or actual errors which may render my remastered edition of Ubuntu buggy ?11:13
cjwatsonara: I'll need to see logs11:38
cjwatsonJemt: they're actual errors (unless they got cleaned up later - see whether 'dpkg --configure -a' still produces errors).  are you sure you should be using 'apt-get upgrade' rather than 'apt-get dist-upgrade'?11:39
aracjwatson, let me get them for you, thanks11:39
Jemtcjwatson: Nope, not sure. I'm actually trying dist-upgrade now. I'll post the result when it's done11:40
JemtHowever, apt-get reported an estimated use of additionally 200 MB disk space. If that's true, upgrading the remastered edition is not an option.11:41
JemtBut let's see how it goes11:41
aracjwatson, http://paste.ubuntu.com/556126/11:44
aracjwatson, I tried in several servers with the same results11:44
cjwatsonJemt: you could remove stale packages afterwards ...11:48
Jemtstale ?11:48
cjwatson(that might be mostly a new kernel)11:48
cjwatsonwell, do you understand why dist-upgrade might produce different results from upgrade?11:48
JemtActually apt-get removed kernel-headers after upgrading, so it released almost 95 MB11:48
Jemtcjwatson: According to the man it has better conflict resolution. Other then that, I don't know much about it11:49
cjwatsonJemt: dist-upgrade is willing to install packages that aren't installed yet, if dependencies require it11:50
cjwatson(and also to remove packages if conflicts require it, but that's less important in this case)11:50
JemtI see11:50
JemtYes11:50
cjwatsonupgrade will only ever install new versions of packages you already have11:50
cjwatsonthis means that if, say, the linux-image-generic package starts depending on linux-image-2.6.35-23-generic rather than linux-image-2.6.35-22-generic, then dist-upgrade will install linux-image-2.6.35-23-generic11:51
cjwatsonhowever, it will generally leave the old packages installed too, unless there's a specific reason to remove them11:51
JemtBut I don't see how "apt-get upgrade" can result in dependency problem. I have never experienced anything like that in a normal environment11:51
cjwatsonthat's what I mean by stale packages11:51
cjwatson'apt-get autoremove' may remove some of those11:51
JemtAh, okay, I see11:52
JemtWell, dist-upgrade is what I want then. Never the less, I still get lots of errors.11:52
cjwatsonit shouldn't normally; I was suggesting dist-upgrade because it's slightly less likely to run into bugs11:52
cjwatsonok, if dist-upgrade does the same, then probably what's happening is that some package's maintainer script is failing11:52
cjwatsonyou haven't given me enough of the output from dpkg for me to be able to tell what that is11:52
JemtI doubt it. I upgraded my host environment without problems11:52
cjwatsoncan I see the full output from 'apt-get upgrade', please?11:53
JemtHang on, placing it on pastebin :)11:53
CIA-47ubiquity: superm1 * r4483 ubiquity/ (bin/oem-config-remove-gtk debian/changelog): Fix oem-config-remove-gtk for changes in AptClient's commit_packages.11:54
JemtUnfortunately I didn't pipe the output to a file, so I don't have the entire output, only what's available in my console buffer: http://pastebin.com/Tz6ysqNr11:55
JemtIf you want, I can run the upgrade again on a clean ISO, to produce all the output11:55
JemtAnd again, I'm performing the upgrade while remastering Ubuntu (I "chrooted" into the squash-fs)11:56
JemtWell, the _extracted_ squash FS11:57
cjwatsonI can't tell from that log - the initial error is before the start11:57
cjwatsonwhat you need to look for is the *first* failure11:58
cjwatsondid you remember to bind-mount /dev, /proc, and /sys into the chroot?11:58
cjwatsonI don't know if all of those will be necessary, but I wouldn't generally want to try to upgrade things without that11:58
cjwatsonara: thanks, this looks rather odd, I'll investigate11:58
cjwatsonara: I might be wrong, but I don't think it can have anything to do with the prior partition table11:59
cjwatsonara: can I have /var/log/partman as well, in case that's necessary?11:59
JemtHm, no, I only bind mount /dev11:59
aracjwatson, sure, do you want me to file a bug against debian-installer?11:59
JemtI'll try bind mounting your suggestions too, and re-run the upgrade. I'll post the result afterwards11:59
aracjwatson, http://paste.ubuntu.com/556135/12:03
JemtI'm away for about 30 minutes. The upgrade should be done then12:03
cjwatsonara: sure, this is pretty bizarre12:08
cjwatsonara: it should be reproducible without any preseeding, I think?12:08
aracjwatson, I haven't tried12:08
* ara files a bug against d-i12:08
CIA-47ubiquity: evand * r4484 trunk/ (156 files in 3 dirs): Update translations from Launchpad.12:11
aracjwatson, bug https://bugs.launchpad.net/ubuntu/+source/debian-installer/+bug/70537712:15
ubot2Launchpad bug 705377 in debian-installer (Ubuntu) "Debian installer prompts for partition type when installing Ubuntu server (affects: 1) (heat: 6)" [Undecided,New]12:15
aracjwatson, thanks!12:15
cjwatsonI'll try it interactively and see what happens12:16
Jemtcjwatson: Alright, update result is now available at http://pastebin.com/rh51MafV12:39
JemtInitial error seems to be related to dbus12:40
Jemt(Line 1615)12:40
cjwatsonit might be worth bind-mounting /var/lock and /var/run as well then.  however, I don't actually know very much about dbus, and it isn't part of the installer12:41
cjwatsonall your other problems seem to arise from that12:42
JemtOkay, I was thinking about trying that too. Wasn't sure whether it could cause other problems though.12:42
cjwatsonI'm not sure about that, I'm afraid12:43
JemtAfter all, the "guest environment" (into which I "chrooted") share resources with the host environment. But it's not a critical computer, so if something breaks, it can easily be restored12:43
JemtI'll give it a try. Thanks  :)12:44
cjwatsonara: broken by superm1's latest change to partman-partitioning12:59
cjwatsonI'll fix it up12:59
cjwatsonsuperm1: perhaps check with me before changing d-i components, sometimes they can be subtle ...13:02
* cjwatson rebases partman-partitioning onto the git migration first13:03
cjwatson(though of course it was my fault that ubiquity broke ...)13:07
CIA-47partman-partitioning: cjwatson * r899 ubuntu/ (debian/changelog lib/disk-label.sh):13:09
CIA-47partman-partitioning: Use 'type' to check for archdetect on PATH, not 'which', which isn't13:09
CIA-47partman-partitioning: available in d-i (LP: #705377).13:09
CIA-47partman-partitioning: cjwatson * r900 ubuntu/debian/ (7 files in 2 dirs): merge from Debian 7913:10
Jemtcjwatson: Thank you very much for your help. Bind mounting /var/run and /var/lock solved the problem. I took a look at UCK (Ubuntu Customization Kit), which seems to do the same (except mounting /var/lock - probably not necessary). However, UCK doesn't do bind mounts - it mounts the resources without the --bind argument. Not sure why. But using --bind ensure the use of the same mount options, so I'll stick to that13:12
Jemtcjwatson: Did you find time to look at the e-mail I sent you ?13:14
cjwatsonnot yet, sorry13:15
JemtNo problem. The problem was not as urgent as the one we just solved :-)13:16
cjwatsonev: looks like you forgot to push your casper change to lp:ubuntu/casper?13:18
cjwatsonev: I've just committed a few other things and got a reject on upload - can you push your branch somewhere and I'll merge and fix it up?13:18
CIA-47partman-partitioning: cjwatson * r901 ubuntu/debian/changelog: releasing version 79ubuntu113:18
aracjwatson, thanks13:30
Jemt"mount --bind /dev/ extracted_squash_fs/dev"  <= Wouldn't you guys expect /dev/pts to be mounted in squash fs too ?13:54
JemtINstalling a package gave me this error: "Can not write log, openpty() failed (/dev/pts not mounted?)"13:54
cjwatsonthat's a warning you can ignore13:56
cjwatson/dev/pts is a separate filesystem; bind-mounting /dev won't bind-mount it too.  See the documentation for mount --bind.  If you want to bind submounts too then you can use --rbind13:56
JemtYes, but I'm still confused. Bind mounting /dev should give me everything in dev, right ?13:56
JemtOh13:56
JemtThank you13:56
evcjwatson: will do.  Sorry about that.  I normally have the branch bound, but I had to recreate it recently.13:57
evcjwatson: lp:~ev/casper/whoops13:58
cjwatsonta13:58
superm1cjwatson, oh sorry. i didn't realize something like which wouldn't be available in d-i.14:00
superm1i'll make sure to test both d-i and ubiquity in the future if changing something that will affect both rather than just ubiquity14:00
cjwatsonyeah, tools are often pretty restricted14:00
cjwatsonthanks14:00
JemtHow is /boot/initrd.img-[version]-generic compressed ? The .img extension puzzles me, since .lz is used in LiveCD/Casper15:39
cjwatsongzip15:40
cjwatson(use the 'file' command)15:40
cjwatsonthe file naming is more or less historical; changing it would have tentacles all over the place so we'd rather not15:40
JemtI see15:42
Jemtfile didn't reveal the compression though15:43
Jemtoutput:  initrd.lz: data15:43
cjwatsonit might not work on lzma - it works on /boot/initrd.img-blah15:45
cjwatsonanyway, .lz is lzma compression15:45
cjwatsonwe recompressed the initrd for the live CD to save space, since the space/time tradeoff is different on live cDs15:46
cjwatsonCDs15:46
JemtMmm, okay. The thing is, I just upgraded the remastered edition, which gave me a more recent kernel. I'd better copy that kernel to LiveCD/Casper, to have it boot with the new kernel. But I should probably use the more recent initrd too15:47
JemtI guess I have to decompress the new initrid.img file and recompress it as lzma15:48
cjwatsonright, although it's certainly possible to copy it to initrd.gz and change isolinux.cfg if you don't want to bother recompressing it15:49
JemtEven better. I can simply copy it to .gz ? Why not keep the .lz extension ?15:49
Jemtsorry, .img extension - why not keep that ?15:50
cjwatsoner, well, it would be pretty confusing to call it .lz if it's gzip-compressed :-)15:50
JemtYes, sorry. I wanted to keep the .img extension :)15:50
cjwatsonI wouldn't advise it - I'm not certain that isolinux can cope with filenames outside the 8.3 format in all cases15:51
JemtOkay, I'll use .gz :)15:51
cjwatsonof course it'll be bigger, I don't know what your space pressures are like15:51
JemtWow, yes - much much bigger - almost 10 times15:52
Jemtlzma sure is effective15:52
JemtWell, I think I can squeeze it to fit on the CD15:53
JemtHm, an alternative solution would be to hold back the kernel during upgrade, to keep the kernel used on the Live CD15:54
JemtI'll take a moment to think it over15:54
cjwatson10x would be pretty unusual; it's only 1.4 times bigger with gzip versus lzma here15:55
cjwatson(real system initrd, not live CD initrd)15:55
JemtMine is 10.762.373 bytes (initrd.img-2.6.35-22-generic)15:56
JemtOdd15:56
JemtDo you have the same version ?15:57
cjwatsonwhat did you do to test that it's 10x smaller when lzma-ing?15:58
Jemt"ls -la /boot" on my laptop vs "ls -la /LiveCDMounted/casper" on the mounted ISO15:59
JemtOh, sorry15:59
JemtI missed a digit in one of them. Sorry, they are almost identical in size16:00
JemtActually the new .img is a bit smaller16:00
cjwatsonalso they don't contain the same things.16:02
JemtYes, probably not16:02
cjwatsonthe initrd is built dynamically16:02
cjwatsonhence why I didn't answer your version question, because it's mostly irrelevant :)16:02
JemtSo it would be different on different computers ??16:02
JemtI didn't see any build processes running during upgrade16:03
cjwatsonyes.  that's what update-initramfs does16:10
JemtHm, I wonder whether it's a good idea to upgrade the kernel then16:10
cjwatsonin particular, casper - the component that deals with the special arrangements needed to boot a live CD - lives in the initramfs16:10
cjwatsonit is not possible to boot a live CD with the same initrd that you use to boot an installed system16:11
JemtOkay, important point :)16:11
JemtI would still like to get the most recent kernel, but will it boot with the "old" initrd.lz on the Live CD ?16:11
cjwatsonno16:12
cjwatsonat least not if the ABI (the 2.6.35-22 part) has changed16:12
JemtUpgrading the kernel sounds like a really bad idea then. I'll keep it back while upgrading then16:12
JemtI'm glad we had this conversation. Otherwise I would have shipped a mess to the school waiting for this customized version of Ubuntu :)16:13
cjwatsonev: do you have any ideas on what to do about bug 562312?  difficult problem ...16:59
* ev reads17:02
evI don't suppose we could do something clever with grub2 like get the kernel and initramfs out of the casper-rw ext3 image?  Forgive me if that's nonsensical, I'm running almost entirely off coffee today.17:05
evcjwatson: ^17:06
cjwatsonev: yes, but relies on switching to grub2, and we ought to do something about older releases if possible17:06
cjwatsonit even bit pitti when he was trying to validate a casper SRU17:07
evsure, but we should do both :)17:07
evas in I'm happy to make the UI change as a backport, but want a better solution than handling it in the UI17:07
cjwatsonactually your suggestion would rely on some kind of union filesystem handling in grub217:08
cjwatsonwhich seems ... likely to be interesting17:08
cjwatsonI suppose you could just conditionalise it or something17:08
evcan you elaborate? My thought was to try to read it from inside casper-rw if it exists, otherwise look in the normal location.17:43
evis that what you mean by conditionalize?17:43
evsorry for the delay, I was trying to play around with aufs, but it just doesn't like me today17:43
cjwatsonthat's what I meant, yes17:44
evare you happy with that? (If so, I can brain dump onto the bug)17:44
cjwatsonit's ok for whenever we manage to get to grub2 on the live images, just don't bank on that being soon :(17:44
evsure, I just want to make it clear to everyone that it's the end goal, rather than the ugly UI living forever17:45
cjwatsonyeah17:45
cjwatsoncould we just cap the amount of space you're allowed to use in casper-rw to allow enough space for a kernel and initrd?17:46
cjwatsonrather than further uglifying the UI17:46
evsure17:46
cjwatsonI realise the size is not entirely constant but it doesn't vary all that much really17:46
evI've updated the bug17:48
cjwatsonthank17:48
cjwatsons17:48
evsure thing17:48
=== charlie-tca is now known as charlie-tca__
=== charlie-tca__ is now known as charlie-tca
Jemtcjwatson: Hello again. You mentioned that update-initramfs builds an image specific to my box. Do you know where I can find more information about this? I wonder how the initrd.lz was build for the Live CD, which must be some sort of "generic" ram disk.20:35
cjwatsonI didn't say it was necessarily *very* specific to your box, just that in principle it can be.20:37
cjwatsondepends on what's in initramfs.conf20:37
cjwatson(/etc/initramfs-tools/)20:37
cjwatsonwe generally use MODULES=most which is pretty generic, but some people have particular needs and strip it down or add stuff or whatever20:37
JemtOkay, I will take a look at the configuration file20:38
JemtI see20:38
cjwatsonthe real difference though is that the casper package is installed when building the initramfs for the live CD.20:38
JemtOh yes, I forgot20:38
cjwatsongenericness isn't likely to be a problem; I just wanted to steer you away from the incorrect notion that there was just one initramfs for any given kernel version20:38
JemtYes, I get it now. I'll have my computer compile a new custom Ubuntu edition tomorrow. It takes well over an hour, and I'm on my way to bed. Hopefully holding back the kernel when upgrading works.20:40
cjwatsonev: could you update umenu/wubi/whatever for the upcoming 10.04.2 point release, please?20:41
JemtI'm off. Thanks for all your help, watson20:53
CIA-47console-setup: cjwatson * r374 ubuntu/debian/ (3 files):20:55
CIA-47console-setup: Correct fix for LP: #634402: explicitly check readability of20:55
CIA-47console-setup: /etc/default/keyboard and /etc/default/console-setup in initramfs hooks,20:55
CIA-47console-setup: rather than trying to guard '.' with '||' which doesn't work20:55
CIA-47console-setup: (LP: #701954).20:55
CIA-47console-setup: cjwatson * r375 ubuntu/debian/changelog: releasing version 1.57ubuntu320:56

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!